Best open-weight LLMs to self-host (2026)

Which strong models ship weights you can run on your own hardware?

Rebuilt from live model data · August 3, 2026

How we ranked

Models with published weights on Hugging Face, ranked by Intelligence Index.

Top 3 for best open-weight llms to self-host

  1. 1

    Kimi K3

    Intelligence index: 57.1 $3.00 in / $15.00 out per 1M 1.0M context 6 providers

    Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at...

    Open weights Vision Tool calling Reasoning Released Jul 16, 2026
  2. 2

    GLM 5.2

    Intelligence index: 51.1 $0.741 in / $2.33 out per 1M 1.0M context 29 providers

    GLM 5.2 is a large-scale reasoning model from Z.ai. It supports text input and output with a 1M-token context window, and is suited for long-horizon agent workflows, project-level software engineering,...

    Open weights Tool calling Reasoning Released Jun 16, 2026
  3. 3

    MiniMax M3

    Intelligence index: 44.4 $0.240 in / $0.960 out per 1M 1.0M context 8 providers

    MiniMax-M3 is a multimodal foundation model from MiniMax. It supports text, image, and video inputs with text output, a 1M-token context window, and is suited for long-horizon agentic work, coding,...

    Open weights Vision Tool calling Reasoning Released May 31, 2026

Full ranking

#ModelIntelligence indexOutput / 1M ContextProvidersWeights
1 Kimi K3 57.1 $15.00 1.0M 6 Open
2 GLM 5.2 51.1 $2.33 1.0M 29 Open
3 MiniMax M3 44.4 $0.960 1.0M 8 Open
4 DeepSeek V4 Pro 44.3 $0.870 1.0M 18 Open
5 Kimi K2.6 44.2 $2.72 262K 21 Open
6 MiMo-V2.5-Pro 42.2 $0.696 1.1M 6 Open
7 Kimi K2.7 Code 41.9 $3.50 262K 13 Open
8 Hy3 preview 41.2 $0.210 262K 1 Open
9 Nex-N2-Pro 41.0 $1.00 262K 2 Open
10 Inkling 40.7 $4.05 1.0M 3 Open
11 DeepSeek V4 Flash 40.3 $0.179 1.0M 21 Open
12 GLM 5.1 40.2 $2.99 205K 19 Open

FAQ

Which strong models ship weights you can run on your own hardware?

Kimi K3 leads for Best open-weight LLMs to self-host with intelligence index 57.1, at $15.00 per 1M output tokens. GLM 5.2 (51.1) and MiniMax M3 (44.4) follow. Models with published weights on Hugging Face, ranked by Intelligence Index.

How is this ranking produced?

Models with published weights on Hugging Face, ranked by Intelligence Index. The list rebuilds from the live model catalogue on every deploy, so a model that launches or changes price appears here without an editor rewriting the page. Last rebuild: August 3, 2026.

What does the top pick cost to run?

Kimi K3 costs about $7.50 for a workload of 1M input + 300K output tokens, based on the cheapest of 6 provider(s).